Author: Nujhat Attar
College: Sinhgad Law College, Pune
To the Point
On 2 July 2026, the Supreme Court set aside two orders, one from the National Company Law Tribunal and one from its appellate counterpart, for a reason nobody had quite seen before in an Indian courtroom: the tribunal below had built its reasoning on case law that simply did not exist. Six precedents cited in the NCLT’s order had been generated by an AI tool, and none of them were real judgments. The Supreme Court did not just correct the record; it laid down a rule for the entire justice system going forward, that any decision resting even partly on fake or hallucinated AI-generated material is not a decision in the eyes of the law.
Use of Legal Jargon
A handful of terms explain what actually happened here. “AI hallucination” is the now well-known tendency of generative AI tools to produce answers that sound completely convincing but are factually invented, including case names, citations, and quoted paragraphs that were never written by any judge. A “precedent” in law means an earlier decided case that binds or guides a court in deciding a similar dispute later; the entire doctrine of stare decisis, meaning courts should stand by earlier decisions, depends on precedents being real and verifiable. The dispute itself arose from a Section 7 application under the Insolvency and Bankruptcy Code, 2016, which allows a financial creditor, here a bank, to approach the NCLT to begin the corporate insolvency resolution process against a company that has defaulted on debt secured by its guarantee. When the Supreme Court held that a tainted decision is “no decision in the eyes of law,” it was invoking the idea of a decision being void ab initio, meaning invalid from the very start, rather than merely voidable or open to being cured later. The Court also flagged this as a matter of professional misconduct for advocates who cite such material without checking it, a term that in Indian law connects directly to disciplinary proceedings under the Advocates Act, 1961, that can be pursued before the Bar Council of India.
The Proof
The case began far away from any AI debate, in a fairly conventional loan default. Jammu and Kashmir Bank had extended a credit facility of roughly 200 crore to Pan India Utilities Distribution Company Ltd, backed by a corporate guarantee from Essel Infraprojects Ltd. When the borrower defaulted and its account turned into a non-performing asset, the bank invoked the guarantee and moved the NCLT in Mumbai under Section 7 of the IBC, seeking to drag Essel Infraprojectsinto insolvency proceedings as the guarantor. The NCLT admitted the application on 28 August 2024, appointed an interim resolution professional, and imposed a moratorium under Section 14 of the Code. The appellant, Pooja RameshSingh, challenged this before the NCLAT and eventually before the Supreme Court through Civil Appeal No. 11950 of 2025. It was only when her counsel dug into the judgment that a strange pattern emerged: several of the precedents the NCLT had relied upon to reject her arguments could not be found in any official law report, and the bank itself later filed an affidavit confirming its own counsel had never cited those judgments at the hearing. In other words, the tribunal appeared to have sourced the fake citations itself, not through a party’s lawyer, which made this arguably the first Indian case where the bench, rather than the bar, introduced hallucinated material into a judgment. A Division Bench of Justices Pamidighantam Sri Narasimha and Alok Aradhe set aside both the NCLT and NCLAT orders, sent the Section 7 application back for a fresh hearing, and used unusually forceful language to describe the danger of unchecked AI use in adjudication, comparing its unchecked spread through the justice system to a toxic industrial leak. The Court also asked the Bar Council of India to examine the broader problem of unverified AI-generated material being cited in litigation, while clarifying that the ruling was not against using AI tools for research as such, only against presenting fabricated output as genuine precedent.
Abstract
This article studies the Supreme Court’s ruling in Pooja RameshSingh v. Jammu and Kashmir Bank Ltd. (2026 INSC 668), the first Indian case in which a tribunal itself, rather than the litigating parties’ counsel, relied on AI-hallucinated case law to decide a matter under the Insolvency and Bankruptcy Code, 2016. It traces how the fabricated precedents entered the NCLT’s reasoning, how the error surfaced on appeal, and how the Supreme Court responded by declaring a zero-tolerance standard, holding that any decision resting even partly on fake AI-generated material is void from inception. The article situates this ruling within a small but growing body of comparative case law from other jurisdictions confronting the same problem, and within established Indian doctrine on natural justice and advocate misconduct, before assessing what practical safeguards courts, tribunals, and lawyers are now expected to build into their research process.
Case Laws
1. Pooja Ramesh Singh v. Jammu and Kashmir Bank Ltd. & Anr., 2026 INSC 668 — the lead case itself; the Supreme Court set aside the NCLT and NCLAT orders after finding the tribunal had relied on six non-existent, AI-hallucinated precedents, declared such tainted decisions void in the eyes of law, and directed the Bar Council of India to examine the issue of unverified AI-generated citations in litigation.
2. Mata v. Avianca Inc. (S.D.N.Y. 2023) — the widely reported United States case where lawyers submitted a brief containing several fictitious judgments generated by ChatGPT; the court imposed sanctions on the attorneys, and the episode is generally treated as the first major judicial confrontation with AI hallucination anywhere in the world.
3. Ayinde v. London Borough of Haringey; Al-Haroun v. Qatar National Bank (England and Wales High Court, 2025) — a joined ruling where the English High Court dealt with lawyers citing fake AI-generated authorities, warning that such conduct could amount to contempt of court or professional misconduct and calling for stronger verification practices before any citation reaches a judgment.
4. Maneka Gandhi v. Union of India (1978) — though unrelated to AI, this foundational ruling on natural justice and fairness in decision-making is routinely invoked whenever the integrity of an adjudicatory process is questioned, since a decision resting on false material denies the affected party the fair hearing that Article 21 guarantees.
5. P.D. Khandekar v. Bar Council of Maharashtra (1984) — a Supreme Court ruling that defines the scope of professional misconduct under the Advocates Act, 1961, a standard that now extends, per the 2026 judgment, to advocates who cite AI-generated case law without independently verifying it before placing it before a court.
Conclusion
What makes this ruling significant is not that a court finally noticed lawyers or clerks leaning on AI shortcuts; legal systems around the world have been bumping into that problem for a couple of years now. What sets the Indian case apart is that the fabricated material appears to have entered the process through the tribunal’s own research, not a litigant’s carelessness, and the Supreme Court responded by refusing to treat that as a minor procedural slip. By declaring that even an indirect trace of fake AI-generated material voids a decision entirely, the Court has effectively told every court and tribunal in the country that convenience cannot come before accuracy. For law students, junior advocates, and increasingly even judges leaning on AI-assisted research tools, the message is straightforward: verify before you cite, because the shortcut that saves an afternoon can unravel an entire judgment months later.
FAQ’s
1.What did the Supreme Court actually decide in this case?
It set aside the NCLT and NCLAT orders because the NCLT had relied on six fake, AI-generated precedents, and sent the underlying insolvency application back for a fresh hearing.
2 .Who introduced the fake case law, the lawyers or the tribunal?
Based on the record, it appears the tribunal itself sourced the fabricated citations, since the bank’s own counsel confirmed those judgments were never argued before it.
3.Does this ruling mean judges and lawyers cannot use AI tools at all?
No. The Court was explicit that it has no issue with the legitimate use of AI for research; the problem is presenting unverified, fabricated output as if it were real precedent.
4.What happens to a judgment found to rest on fake AI citations?
The Supreme Court held that such a judgment is void in the eyes of law and must be set aside, even if the fake material only had an indirect bearing on the outcome.
5.Can a lawyer get in trouble for citing AI-hallucinated cases?
Yes. The Court described this as professional misconduct on the part of advocates and has asked the Bar Council of India to examine the issue, which could lead to disciplinary consequences.
6.Is India the only country dealing with this problem?
No. Courts in the United States and England have dealt with similar episodes of AI-fabricated case law being cited in filings, though this appears to be the first Indian instance traced back to a tribunal’s own research rather than a lawyer’s filing.
